G
G
E
E
N
N
E
E
T
T
I
I
C
C
A
A
N
N
D
D
P
P
R
R
O
O
C
C
E
E
S
S
S
S
E
E
N
N
G
G
I
I
N
N
E
E
E
E
R
R
I
I
N
N
G
G
F
F
O
O
R
R
T
T
H
H
E
E
R
R
E
E
C
C
O
O
M
M
B
B
I
I
N
N
A
A
N
N
T
T
P
P
R
R
O
O
D
D
U
U
C
C
T
T
I
I
O
O
N
N
O
O
F
F
P
P
R
R
O
O
T
T
E
E
I
I
N
N
T
T
H
H
E
E
R
R
A
A
P
P
E
E
U
U
T
T
I
I
C
C
S
S
F
F
O
O
R
R
T
T
H
H
E
E
T
T
R
R
E
E
A
A
T
T
M
M
E
E
N
N
T
T
O
O
F
F
CN
C
NS
S
D
D
I
I
S
S
O
O
R
R
D
D
E
E
R
R
S
S
Andrea Colarusso
Dottorato in Biotecnologie – XXXI ciclo
Dottorato in Biotecnologie – XXXI ciclo Università di Napoli Federico II
G
G
E
E
N
N
E
E
T
T
I
I
C
C
A
A
N
N
D
D
P
P
R
R
O
O
C
C
E
E
S
S
S
S
E
E
N
N
G
G
I
I
N
N
E
E
E
E
R
R
I
I
N
N
G
G
F
F
O
O
R
R
T
T
H
H
E
E
R
R
E
E
C
C
O
O
M
M
B
B
I
I
N
N
A
A
N
N
T
T
P
P
R
R
O
O
D
D
U
U
C
C
T
T
I
I
O
O
N
N
O
O
F
F
P
P
R
R
O
O
T
T
E
E
I
I
N
N
T
T
H
H
E
E
R
R
A
A
P
P
E
E
U
U
T
T
I
I
C
C
S
S
F
F
O
O
R
R
T
T
H
H
E
E
T
T
R
R
E
E
A
A
T
T
M
M
E
E
N
N
T
T
O
O
F
F
CN
C
NS
S
D
D
I
I
S
S
O
O
R
R
D
D
E
E
R
R
S
S
Andrea Colarusso
Dottorando:
Andrea Colarusso
Relatore:
Prof.ssa Maria Luisa Tutino
Coordinatore:
Prof. Giovanni Sannia
Ad Elettra Yvonne,
Index
Acknowledgments 1
Riassunto 2
Summary 7
Chapter 1: CNF1 production and purification for its functional and
therapeutic characterization 8
1.1 Introduction 8
1.1.1 CNF1 toxin as a prototypical virulence factor targeting Rho GTPases 8 1.1.2 CNF1 toxin as a drug candidate for the treatment of CNS disorders 10
1.2 Aim of the study 12
1.3 Results 12
1.4 Discussion 23
1.5 References 25
Chapter 2: Recombinant production of a protein involved in human
cerebral development in Pseudoalteromonas haloplanktis TAC125 30
2.1 Introduction 30
2.1.1 Variant B: a paramount protein in brain development 30
2.1.2 Pseudoalteromonas haloplanktis TAC125: an exotic bacterium for the
recombinant production of human proteins 30
2.2 Aim of the study 33
2.6 References 34
1
Acknowledgments
This PhD project and my fellowship were funded by the Italian parent association “La
Fabbrica dei Sogni 2 – New Developments for Rett Syndrome” and by the company
Amicus Therapeutics Inc. I thank both for having supported my research.
I am especially grateful to Prof Maria Luisa Tutino for having allowed me to work in her team on very stimulating projects. I thank her for her guidance and her constant interest in my work and my results.
I also wish to thank Prof Ermenegilda Parrilli for her critical attitude towards each experiment and for her valuable technical advices.
Finally, I would like to thank Gennaro Antonio Apuzzo, Marzia Calvanese and Concetta Lauro for their technical and decisional support in the development of part of my work.
2
Ingegnerizzazione genetica e di processo per la produzione ricombinante di proteine terapeutiche per il trattamento di disordini del SNC
Riassunto
Introduzione
I disordini neurologici costituiscono una delle cause principali di disabilità al mondo con enormi conseguenze sulla qualità media della vita delle persone affette e di quella delle persone loro vicine (Feigin et al., 2017). In particolare, con l’incremento delle aspettative di vita, patologie quali il morbo di Alzheimer e quelle associate a varie forme di demenza sono da anni rientrate fra i disordini con maggior prevalenza in questo campo (Feigin et al., 2017). Nonostante questo fardello sia diventato via via più pressante col tempo, ad oggi l’industria farmaceutica globale non è riuscita a proporre efficaci misure risolutive per l’ampio e diversificato spettro di patologie in questo settore. Come dimostra una recente analisi statistica, infatti, nel decennio 2006 – 2015 solo l’8.4% dei potenziali farmaci proposti nell’ambito neurologico ha incontrato un favorevole parere dell’FDA (Hay et al., 2014). Se si considerano poi i soli candidati rientranti nella nomenclatura “Grandi Molecole” in tutti i settori, si denota che il 13.2% di essi ha una possibilità di essere accettato come reale farmaco. Al di là di misure di controllo talvolta eccessivamente restrittive, un tasso di successo così basso scaturisce sovente da un’inefficiente valutazione delle reali potenzialità dei composti proposti per lo sviluppo di farmaci in fase preclinica. Date, infatti, le scarse possibilità di approvazione finale da parte degli organi competenti, viene logico ritenere che solo quelle molecole che fin dalle prime fasi di sviluppo dimostrano un’inconfutabile efficacia e un chiaro e sicuro meccanismo di azione meritino di essere portate avanti nel processo. Un’eccezione in questo quadro sfavorevole è costituita da quelle proteine umane che generalmente vengono prodotte ed utilizzate in terapie sostitutive per il trattamento di patologie monogeniche. Dato il basso grado di novità di queste molecole, in genere la probabilità di essere rapidamente accettate alla fine del processo di valutazione del farmaco è più alta (Gorzelany and De Souza, 2013). In questi casi sono le procedure di produzione e formulazione finale del prodotto stesso le fasi più critiche dell’intero iter di sviluppo, data la natura delicata delle molecole in oggetto (Saccardo, Corchero and Ferrer-Miralles, 2016).
In questo lavoro sono descritte le strategie e le procedure sperimentali impiegate nel corso della produzione ricombinante di due proteine con potenziali applicazioni nel trattamento di patologie del sistema nervoso centrale (SNC). Entrambe hanno imposto diverse sfide inerenti ad alcuni degli aspetti sopra accennati e le varie soluzioni proposte per superarle sono riportate.
Capitolo 1: Produzione e purificazione di CNF1 per la sua caratterizzazione funzionale e terapeutica
La tossina CNF1 (Cytotoxic Necrotizing Factor 1) è una proteina di 1014 amminoacidi secreta da alcuni ceppi patogeni di Escherichia coli nel corso di infezioni urinarie, gastrointestinali e meningiti. Il meccanismo di azione di tale tossina prevede l’interazione con due recettori di membrana largamente diffusi in vari tessuti dell’organismo umano. Dopodiché, essa è internalizzata nella cellula mediante endocitosi ed il suo dominio catalitico è rilasciato nel citosol dopo uno specifico taglio proteolitico. In questo compartimento cellulare tale dominio catalizza una reazione di deammidazione a carico di un residuo di glutammina delle Rho GTPasi RhoA, Rac1
3
e Cdc42. Complessivamente, tale modifica irreversibile induce l’attivazione costitutiva e la successiva degradazione proteasomale di questi regolatori cellulari, così da alterarne i fisiologici livelli cellulari. Questo processo è sfruttato dai batteri patogeni per interferire con il sistema immunitario dell’ospite e per facilitare la loro penetrazione attraverso i tessuti cellulari. L’attività delle Rho GTPasi è, infatti, intrinsecamente legata alla produzione di citochine proinfiammatorie, alla regolazione dell’organizzazione del citoscheletro di actina, alla forma, alla motilità ed alla vitalità cellulare. Sulla base dell’analisi dei peculiari fenotipi indotti dalla trasduzione di CNF1 all’interno del SNC, un gruppo italiano ha proposto l’impiego di tale tossina per il trattamento sintomatologico di patologie e disordini neuronali, quali la malattia di Alzheimer, il morbo di Parkinson, la sindrome di Rett e l’epilessia. Nonostante gli interessanti studi preliminari condotti su modelli murini delle summenzionate patologie, un’estensiva e robusta analisi preclinica che attesti la sicurezza e l’efficacia di un potenziale farmaco fondato su questa molecola non è stato ancora condotto. Per di più, sia in prospettiva di validazione clinica che di eventuale produzione industriale è essenziale verificare la possibilità di ottenere il bioprodotto di interesse a livelli accettabili in termini di rese, purezza e stabilità. I protocolli di espressione e purificazione di CNF1 riportati in letteratura sono abbastanza deludenti in tal senso. In taluni casi è ravvisabile una notevole contaminazione del campione finale da parte di frammenti di degradazione della proteina in esame o di altre proteine provenienti dall’ospite di espressione. Questo problema è stato arginato facendo ricorso a procedure di purificazione particolarmente complesse e lunghe che constano anche di oltre dieci passaggi consecutivi con un notevole impatto sulle rese finali. Anche quando sono stati adoperati metodi di purificazione più semplici, come una singola cromatografia di affinità, non è stato comunque possibile raccogliere una quantità di proteina ricombinante superiore all’ordine del centinaio di µg per litro di coltura. Infine, nessun dato circa la stabilità nel tempo della proteina è mai stato riportato.
Lo scopo principale di questo lavoro è consistito, dunque, nella definizione di una procedura semplice e riproducibile per l’espressione ricombinante e la purificazione di CNF1 raggiungendo standard quantitativi e qualitativi prossimi a quelli di un processo industriale. La proteina così ottenuta sarebbe stata poi impiegata in una successiva caratterizzazione in vitro ed in vivo in un’ottica di una sua valutazione preclinica. I dati così ottenuti sarebbero stati essenziali per valutare se procedere o meno con lo sviluppo di un potenziale farmaco per il trattamento dei disordini del SNC.
Per il raggiungimento degli obiettivi prefissati, la proteina in esame è stata prodotta per via ricombinante nel citosol di un ceppo commerciale di E. coli BL21(DE3) inducibile da IPTG, fondendola ad un tag di 8 istidine, così da ottenere il costrutto denominato CNF1-H8. La definizione delle migliori condizioni di crescita e di espressione è stata essenziale per snellire la successiva procedura di purificazione e per preservare la proteina in forma prevalentemente integra. Nella fattispecie, un’induzione condotta ad una temperatura bassa per questo ospite mesofilo (15 °C) ha costituito l’espediente più utile per arginare la “violenza” tipica del sistema di induzione usato (basato sulla T7 RNA polimerasi dei sistemi pET) e per minimizzare i processi degradativi maggiormente ravvisabili a temperature più alte (da 20 °C in su). La buona qualità della produzione e i discreti livelli di accumulo registrati in fase di espressione (ravvisabili anche solo per SDS-PAGE), hanno consentito di ottenere circa 7 - 8 mg di proteina pura per litro di coltura (> 95 % del contenuto proteico totale) alla fine dei passaggi cromatografici impiegati. In particolar modo,
4
l’accoppiamento di una cromatografia di affinità (IMAC) con una ad esclusione molecolare è stato sufficiente per raggiungere gli standard produttivi prefissati. L’utilizzo del secondo passaggio cromatografico è stato essenziale non solo per incrementare significativamente la purezza del campione finale, ma anche per diminuirne l’eterogeneità in termini di aggregazione. Come dimostrato dalle analisi condotte ricorrendo al Dynamic Light Scattering (DLS), infatti, solo con l’eliminazione finale degli aggregati proteici è stato possibile ottenere un prodotto relativamente omogeneo (Pdi 0.275) preservabile a 4 °C per tempi lunghi. Le analisi per DLS insieme ai saggi di attività condotti utilizzando la proteina RhoA come substrato sono state anche utili per giungere alla definizione di una formulazione stabile nel tempo. In particolar modo, per poter essere conservato a concentrazioni piuttosto alte (10 mg/mL), CNF1-H8 ha richiesto l’utilizzo di un tampone con una forza ionica relativamente alta (almeno 150 mM NaCl) e di 15% v/v glicerolo come co-solvente. Fra i due fattori il ricorso ad una elevata concentrazione salina durante la conservazione è risultato essere il più determinante sia per la preservazione dell’attività enzimatica che per il mantenimento di una condizione di quasi monodispersione nel tempo. Per di più, i saggi di intossicazione in vitro eseguiti su colture di cellule HEp-2 hanno confermato che alla fine della procedura di purificazione la proteina era correttamente strutturata per poter essere internalizzata e processata nelle cellule eucariotiche ed esplicare la sua attività catalitica. Seppure alcuni componenti della soluzione tampone di preservazione avrebbero potuto costituire di per se stessi una causa di tossicità, si deve considerare che le preparazioni erano così concentrate che la diluzione nella soluzione del saggio finale ha reso ininfluente il loro contributo. Infine, per la prima volta è stata condotta una generica caratterizzazione strutturale della tossina CNF1-H8 nella sua interezza, ricorrendo ad un’indagine di dicroismo circolare (CD). Lo spettro ottenuto nel lontano UV ha confermato i dati pregressi di predizione di strutture secondarie, indicando che circa il 50% della proteina è molto flessibile e questo probabilmente costituisce il maggior scoglio per la definizione della struttura tridimensionale dell’intera tossina. Differentemente dalle analisi per DLS, gli studi di stabilità termica condotti con il CD hanno inoltre suggerito che la proteina potrebbe essere più stabile ad un pH lievemente acido (pH 6.6), dando ulteriori spunti per successive condizioni da saggiare per la preservazione della molecola.
Nonostante le buone performance raggiunte nel processo produttivo, lo sviluppo di CNF1-H8 od un suo derivato come farmaco non è stato portato avanti. In assenza di un’esauriente e puntuale validazione in termini di Target Engagement, Proof of
Mechanism, Proof of Principle e Proof of Concept, proseguire il progetto avrebbe
costituito un notevole rischio in termini economici e di opportunità. Nonostante tutto, le comprovate riproducibilità e semplicità dei protocolli di produzione impiegati potrebbero costituire una buona base per l’ulteriore caratterizzazione della proteina in chiave tossicologica oppure per il suo impiego come immuno-adiuvante.
Capitolo 2: Produzione ricombinante di una proteina coinvolta nello sviluppo cerebrale in Pseudoalteromonas haloplaktis TAC125
La seconda parte di questo lavoro di tesi ha riguardato la produzione di una proteina umana (in questo testo denominata Variant B) per la sua caratterizzazione ed il suo impiego in una terapia sostitutiva. Le informazioni relative al funzionamento di questa molecola sono ancora piuttosto lacunose anche se un’intensa ricerca è in corso in vari centri per studiarne le implicazioni nello sviluppo cerebrale. Infatti, seppure diverse isoforme di Variant B sono espresse in molteplici tessuti cellulari, è evidente
5
che la sua subnormale espressione nel SNC comporta i più gravi deficit fenotipici durante lo sviluppo infantile. Studi preclinici in modelli murini hanno dimostrato che una terapia proteica sostitutiva può effettivamente ripristinare un fenotipo sano in soggetti affetti da mutazioni aberranti a carico del gene variant B. Nonostante ciò, una cura non è ancora praticabile date le grandi difficoltà nel produrre la proteina per via ricombinante con le piattaforme di espressione più canoniche.
Il gruppo di ricerca in cui ho condotto questo lavoro di tesi si è quindi proposto di impiegare un batterio non convenzionale per tentare l’espressione di Variant B. Si tratta di Pseudoalteromonas haloplanktis TAC125 (PhTAC125), un γ-proteobatterio antartico marino ampiamente caratterizzato da un punto di vista genetico e metabolico e la cui utilità biotecnologica è stata dimostrata in vari ambiti. In particolar modo, mediante la fusione dell’origine di replicazione del plasmide endogeno pMtBL di PhTAC125 a frammenti di plasmidi di E. coli è stata sviluppata una serie di vettori di clonaggio e di espressione da poter utilizzare nel batterio polare. In questo modo, più proteine eucariotiche sono state sintetizzate per via ricombinante in PhTAC125 a partire da questi vettori navetta dimostrando il loro efficace impiego per la produzione di proteine “difficili”.
7
Summary
Neurological disorders constitute the major cause of disability adjusted-life years (DALYs). Alzheimer’s disease and other dementias are included in the group of the most prevalent disorders in this field (Feigin et al., 2017). Nevertheless, the urgency of the treatment of these pathologies is not met by efficacious drug development, as indicated by a recent statistical analysis measuring the likelihood of approval of new drugs by disease area (Hay et al., 2014). In neurology only 8.4% of the candidate drugs have been approved in the 2006 – 2015 decade and the drugs categorized as “Large Molecules” were characterized by 13.2% rate of success (Hay et al., 2014). This statistical analysis would suggest that only those molecules proving robust Proof of Mechanism (PoM) and Proof of Principle (PoP) during their early development deserve the risk for further development. On the other hand, proteins used in protein replacement therapies (PRTs) constitute an exception in this scenario as the iter for their approval is generally more straightforward (Gorzelany and De Souza, 2013). In this case the main hurdles involving the drug development directly coincide with the production, the purification and the stable formulation of the final product rather than the assessment of its efficacy or toxicity (Saccardo, Corchero and Ferrer-Miralles, 2016).
In the present document we describe the approaches followed in the implementation of biotechnological processes for the production of two different proteins with potential applications in the treatment of central nervous system (CNS) disorders. In the first chapter we present the establishment of a simple and efficient pipeline for the E. coli recombinant expression and purification of the bacterial toxin named CNF1. This 114 kDa protein is involved in a series of infectious diseases (Ho et al., 2018), but it has also been demonstrated to be promising in the treatment of severe neurological pathologies like Alzheimer’s disease, Parkinson disease, Rett syndrome and epilepsy (Maroccia et al., 2018). Nevertheless, during the development of the project we decided not to pursue any further attempt in the clinical development of CNF1 as a drug because of the lack of robust, clear and completely demonstrated PoM and PoP. However, the proposed procedure for the purification and final formulation of the product outperforms the others reported in the literature in terms of yield, purity and stability and it can be easily employed in the future for further structural and functional analyses in toxicological and immunological perspectives. The reproducibility of the entire pipeline has been demonstrated repeating the production and purification protocols dozens of times at intervals of several months. Moreover, the stability of the final product was routinely ascertained using SDS-PAGE, size-exclusion chromatography, DLS and activity assays.
In the second chapter of this thesis we describe the employment of
Pseudoalteromonas haloplanktis TAC125 in the production of a human protein to be
used in a PRT. Although the quality of the product achievable in this host seemed better than the one previously obtained with other expression systems, the overall yields remained low. Slight improvements in this sense were achieved by the genetic engineering of the coding sequence of the target protein and by the implementation of alternative expression plasmids. Nevertheless, further studies demonstrated that the whole expression platform – the host and the plasmids – are affected by imperfections and bottlenecks whose correction is pivotal for a satisfying recombinant production. This kind of issues is typical of unconventional and less explored recombinant bacteria, but there are several examples in the literature about how they can be systematically overcome. Hence, a series of measures to be taken for the improvement of this microbial factory are proposed.
8
Chapter 1: CNF1 production and purification for its functional and therapeutic characterization
1.1 Introduction
1.1.1 CNF1 toxin as a prototypical virulence factor targeting Rho GTPases
“Cytotoxic Necrotizing Factor 1” takes its name from the observation that its inoculation in eukaryotic cell cultures induces cellular multinucleation and its subcutaneous injection provokes the appearance of skin necrotic lesions (Caprioli et
al., 1983, 1984). Over the years, CNF1 has become the most studied representative
of a growing family of tightly similar toxins that includes at least nine close homologs secreted by different bacteria for the intoxication of different hosts (Ho et al., 2018). However, CNF1 is exclusively produced by some pathogenic Escherichia coli strains mainly involved in urinary tract infections, intestinal infections and meningitis in humans (Falbo et al., 1993; Khan et al., 2002). In such disparate tissues this bacterial protein interferes with the functioning of Rho GTPases, master regulators of cellular homeostasis, with the overall effect of increasing the colonizing properties of the producing strains (Fiorentini et al., 1997; Flatau et al., 1997; Schmidt et al., 1997; Lerm et al., 1999; Doye et al., 2002). Rho GTPases are indeed targets of various bacterial effectors due to their involvement in a series of processes which can be exploited by pathogens to potentiate their invasion (Boquet and Lemichez, 2003; Galan, 2009; Lemichez and Aktories, 2013). Depending on the mechanism of action, the breach of Rho GTPases regulated processes by toxins can lead to: 1) disruption of host barriers; 2) invasion of non-phagocytic cells; 3) interference with the immune system. Actually, thanks to its sneaky mode of action, CNF1 acts on all the three abovementioned phenomena.
CNF1 is a 1014 aa AB toxin which, after secretion, enters mammal cells by the binding with the laminin receptor precursor 67LR together with the Lutheran adhesion glycoprotein/basal cell adhesion molecule (Lu/BCAM, Lemichez et al., 1997; Fabbri, Gauthier and Boquet, 1999; Chung et al., 2003; Kim, Chung and Kim, 2005b; McNichol et al., 2007; Piteau et al., 2014). Then it triggers the constitutive activation and induces the proteasomal degradation of Rho GTPases by deamidation of a glutamine residue (Gln 63 in RhoA, Gln 61 in Cdc42 and Rac1, Fiorentini et al., 1997; Flatau et al., 1997; Schmidt et al., 1997; Lerm et al., 1999; Doye et al., 2002). The mechanism of internalization of this toxin is quite complex and requires its binding with both the receptors, its transport via endocytosis and the release of the cleaved C-terminal domain in the cytosol after an acidic-dependent insertion of two hydrophobic regions in the endosomal membrane (Contamin et al., 2000; Pei, Doye and Boquet, 2001; Doye et al., 2002, Figure 1.1).
9
Figure 1.1. Schematic representation of CNF1 functional domains and its mechanism of action. CNF1 is a 1014 aa protein belonging to the AB toxin category. At least two
recognition sequences have been identified to interact with laminin receptors: residues 1-342 binding with p67-LR and residues 709-730 with Lu/BCAM. After receptor-mediated endocytosis (1) and endosome acidification, the insertion of two hydrophobic helices (H1 and H2) in the endosomal membrane allows the cleavage (2) and the release in the cell cytosol of a 55 kDa fragment encompassing the catalytic domain (deamidase). Here the C-terminal domain deamidates Rho GTPases specific residues (4), triggering a series of cellular effects summed up in the figure. Taken from Ho et al., 2018.
Although the process of activation and the following proteolysis of the effector proteins might seem a contradiction in the perspective of a bacterial invasion, this temporal switch makes sense in the context of a mechanism of defense exploited both by the host and the pathogen for their preservation. One of the main effects of Rho GTPases activation involves a wide actin cytoskeletal reorganization that is translated in the formation of different ultrastructures, depending on the predominantly activated protein: stress fibers (RhoA), lamellipodia (Rac1), filopodia (Cdc42) (Boquet and Lemichez, 2003). In any case such rearrangements deeply affect cellular polarization, motility and the integrity of cellular junctions, leading to an increased permeabilization of endothelial barriers (Lemichez and Aktories, 2013). At the same time (Fiorentini et al., 2001), the induction of the formation of extracellular protrusions facilitates unspecific phagocytosis and macropynocitosis of extracellular material by unspecialized cells. Moreover, CNF1 induces cellular proliferation thanks to an altered expression of Bcl-2 family proteins (Fiorentini et al., 1998) via the AKT/NF-kB regulatory pathway (Miraglia et al., 2007). All in all, these cellular responses are used to disrupt endothelial barriers and facilitate intracellular migration of the pathogens. Nevertheless, an uncontrolled Rho GTPases hyperactivation would be detrimental both for the host and the pathogen because it would provoke excessive tissues alterations. As G proteins are canonical target of many bacterial
10
effectors and their dysregulation leads to important alterations in cellular homeostasis, mammals have evolved a defense mechanism, through which a strong immune response is triggered when Rho proteins are abnormally activated (Boquet and Lemichez, 2003). Nevertheless, an uncontrolled defense can become harmful even for the host, as it might lead to damages to its own “hyperactivated” tissues. That is why, eukaryotic cells recognize postranslationally modified and constitutively activated Rho GTPases as aberrant proteins and shunt them towards ubiquitination and degradation, so that the violent immune alarm is shut down (Lemonnier, Landraud and Lemichez, 2007). As a result of this sophisticated mechanism of host-pathogen interaction, CNF1 intoxication often leads to a mild activation of Rho proteins (among which RhoA seems the predominant) that triggers a controlled immune response which is neither harmful for the host nor extremely efficient to defeat the pathogen (Munro et al., 2004). The systematic exploitation of these complex endogenous regulatory mechanisms by pathogens to bypass the immune response is further confirmed by the existence of redundant and counteracting bacterial effectors, like hemolysin whose function, among the others, seems to consist in the reduction of the production of cytokines induced by CNF1 action (Diabate et al., 2015). In other words, even if during the evolution these pathogens became detectable by the immune system because of CNF1 action, they managed to preserve this useful toxin by evading the host defense both releasing damage control co-effectors and exploiting the repair mechanisms of the host itself.
This general overview of the mode of action of this family of toxins highlights how the subtle alteration of important homeostatic processes resulting from host-pathogen crosstalk are crucial both for bacterial survival and for the preservation of the host. Paradoxically, if properly exploited, these same processes might be used to produce positive effects in the host, as explained in the following paragraph.
1.1.2 CNF1 toxin as a drug candidate for the treatment of CNS disorders
As discussed in the previous section, the main direct effects of the action of CNF1 on Rho GTPases consist of cytoskeletal reorganization (Boquet and Lemichez, 2003) and the activation of phosphorylation regulatory pathways influencing many processes, including mitochondrial homeostasis (Miraglia et al., 2007). Some both chronic and acute diseases affecting the brain are characterized by common traits, including neuronal atrophy, poor plasticity, and bioenergetics issues (Beal, 2004; Wilkins et al., 2014; Carvalho et al., 2015). Hence, an Italian research group working at Istituto Superiore di Sanità proposed the use of CNF1 to relieve the symptoms of such diseases, proving its efficacy first in the treatment of Rett syndrome (De Filippis
et al., 2012; De Filippis, Valenti, Chiodi, et al., 2015; De Filippis, Valenti, De Bari, et al., 2015) and, then, in Alzheimer’s disease (Loizzo et al., 2013) and Parkinson’s
disease (Musilli et al., 2016) in mice models. The molecular reasons that are at the basis of these explorative and preliminary results are effectively summarized in a recent review (Maroccia et al., 2018) and in Figure 1.2. Basically, the cytoskeletal reorganization induced by activated Rho GTPases is capable of reverting astrocytes atrophy and to restore their supportive role on neurons (Malchiodi-Albedi et al., 2012). On the other side, the activation of the cAMP/PKA phosphorylation cascade directly influences mitochondria functioning, both by inhibiting their fission and by directly regulating proteins involved in the respiratory chain (Travaglione et al., 2014). This same last process proved to be promising in the treatment of epilepsy (Fabbri et
11
capability to alter the normal cell cycle progression was demonstrated to be successful in the treatment of glioma (Vannini et al., 2014).
Figure 1.2. Schematic outline of CNF1 therapeutic features in animal models recapitulating CNS diseases.
Rho GTPases functioning is perturbed by CNF1 catalyzed deamidation and
proteasome degradation. This is
translated in the induction of a profound actin reorganization of the intoxicated cells (left side) and in the triggering of a phosphorylation cascade started by
cAMP accumulation (right side).
Cytoskeletal alteration is useful to increase synaptic plasticity and revert astrocytes atrophy in CNS diseases on one side, and to block the proliferation of cancer cells on the other. The activation of the PKA phosphorylation
cascade ameliorates the cellular
bioenergetics, allowing both an increase of ATP production and a reduction of ROS. This result is mainly due to the
inactivation of the mitochondrial
profission DrpI protein. Taken from Maroccia et al., 2018.
All the above-mentioned proof-of-concept studies that have been carried out in the last decade show how CNF1 action on Rho GTPases is useful to interfere with some CNS diseases and brain functioning in general. Nevertheless, little has been done to investigate its real applicability as a therapeutic drug so far. This section of the present document aims to give a small contribution in this sense, both by providing a protocol for the production and purification of the toxin and some considerations about its use.
12
1.2 Aim of the study
This project started as a collaboration with Dr. C. Fiorentini and A. Fabbri working at Istituto Superiore di Sanità and it aimed to evaluate the feasibility of the use of CNF1 toxin as a protein drug in the treatment of CNS disorders. It was supposed to consist of four main phases:
1) The establishment of a simple procedure for the recombinant production and purification of the wt protein.
2) The evaluation of the fundamental pharmacodynamics (PD) and pharmacokinetics (PK) parameters using different routes of administration, like bioavailability, distribution, efficacy, toxicity, etc.
3) The rational design of CNF1 variants suitable for systemic administration on the base of the PD and PK preliminary data and of the already existing structural/functional information.
4) Characterization of the engineered CNF1 variants.
Because of the lack of collaboration of our partners, the project never moved further than the first step.
1.3 Results
The results of this section are reported in the following published article:
Colarusso, A., Caterino, M., Fabbri, A., Fiorentini, C., Vergara, A., Sica, F., Parrilli, E. and Tutino, M. L. (2018), High yield purification and first structural characterization of the full‐length bacterial toxin CNF1. Biotechnol Progress, 34: 150-159. doi:10.1002/btpr.2574.
High Yield Purification and First Structural Characterization of the Full-Length
Bacterial Toxin CNF1
Andrea Colarusso
Department of Chemical Sciences, University of Naples Federico II, Complesso Universitario Monte Sant’Angelo, Via Cinthia, Napoli, 80126, Italy
Marco Caterino
Department of Chemical Sciences, University of Naples Federico II, Complesso Universitario Monte Sant’Angelo, Via Cinthia, Napoli, 80126, Italy
Alessia Fabbri
Italian Center for Global Health, Istituto Superiore di Sanita, Viale Regina Elena, 299, Roma, 00161, Italy
Carla Fiorentini
Italian Center for Global Health, Istituto Superiore di Sanita, Viale Regina Elena, 299, Roma, 00161, Italy
Alessandro Vergara
Department of Chemical Sciences, University of Naples Federico II, Complesso Universitario Monte Sant’Angelo, Via Cinthia, Napoli, 80126, Italy
CEINGE Biotecnologie Avanzate scarl, Via G. Salvatore, Napoli, 80100, Italy
Institute of Biostructures and Biomaging, CNR, Napoli, Italia Via Mezzocannone 16, Napoli, 80134, Italy
Filomena Sica
Department of Chemical Sciences, University of Naples Federico II, Complesso Universitario Monte Sant’Angelo, Via Cinthia, Napoli, 80126, Italy
Institute of Biostructures and Biomaging, CNR, Napoli, Italia Via Mezzocannone 16, Napoli, 80134, Italy
Ermenegilda Parrilli
Department of Chemical Sciences, University of Naples Federico II, Complesso Universitario Monte Sant’Angelo, Via Cinthia, Napoli, 80126, Italy
Maria Luisa Tutino
Department of Chemical Sciences, University of Naples Federico II, Complesso Universitario Monte Sant’Angelo, Via Cinthia, Napoli, 80126, Italy
DOI 10.1002/btpr.2574
Published online 00 Month 2017 in Wiley Online Library (wileyonlinelibrary.com)
The Cytotoxic Necrotizing Factor 1 (CNF1) is a bacterial toxin secreted by certain Escherichia colistrains causing severe pathologies, making it a protein of pivotal interest in toxicology. In parallel, the CNF1 capability to influence important neuronal processes, like neuronal arborization, astrocytic support, and efficient ATP production, has been efficiently used in the treatment of neurological diseases, making it a promising candidate for therapy. Nonetheless, there are still some unsolved issues about the CNF1 mechanism of action and structuration probably caused by the difficulty to achieve sufficient amounts of the full-length protein for further studies. Here, we propose an efficient strategy for the production and purification of this toxin as a his-tagged recombinant protein from E. coli extracts (CNF1-H8). CNF1-H8 was expressed at the low temperature of 158C to diminish its charac-teristic degradation. Then, its purification was achieved using an immobilized metal affinity chromatography (IMAC) and a size exclusion chromatography so as to collect up to 8 mg of protein per liter of culture in a highly pure form. Routine dynamic light scattering (DLS) experiments showed that the recombinant protein preparations were homogeneous and pre-served this state for a long time. Furthermore, CNF1-H8 functionality was confirmed by test-ing its activity on purified RhoA and on HEp-2 cultured cells. Finally, a first structural characterization of the full-length toxin in terms of secondary structure and thermal stability was performed by circular dichroism (CD). These studies demonstrate that our system can
Correspondence concerning this article should be addressed to Maria Luisa Tutino at tutino@unina.it.
be used to produce high quantities of pure recombinant protein for a detailed structural analysis.VC 2017 American Institute of Chemical Engineers Biotechnol. Prog., 000:000–000,
2017
Keywords: CNF1, recombinant protein production, protein purification, circular dichroism, sec-ondary structure analysis
Introduction
The Cytotoxic Necrotizing Factor 1 (CNF1) is a 1014 aa toxin produced and secreted by some Escherichia coli pathogenic strains principally involved in urinary tract infections and neo-natal meningitis.1,2Although the understanding of CNF1 direct role in bacterial infection is complicated by the overlap with the action of other bacterial effectors,3–5 it has been extensively demonstrated that this toxin triggers the constitutive activation and induces the subsequent proteasomal degradation of Rho GTPases by deamidation of a glutamine residue (Gln63 in RhoA, Gln61 in Cdc42 and Rac).6–10The main effects of this process include the activation of a controlled immune response in the host11and a wide reorganization of the actin cytoskeleton of the infected cell which is exploited for the bacterial uptake.12,13 CNF1 capability to directly affect Rho GTPases action and indirectly actin assembly and cellular bioenergetics has recently been exploited in the treatment of some severe neu-rological diseases, including Alzheimer’s disease, Parkinson’s disease and Rett syndrome.14–22
Several studies have pointed out the general structuration of CNF1 as an AB toxin, presenting two different receptor-recognition sites (an N-terminal 37LRP/67LR binding domain and a C-terminal Lu/BCAM binding domain) and a C-terminal catalytic domain catalyzing Rho deamidation.23–28The mecha-nism of internalization of this toxin is quite complex and requires its binding with both receptors, its transport via endo-cytosis and the release of the cleaved C-terminal domain in the cytosol after an acidic-dependent insertion of two hydrophobic regions in the endosomal membrane.28–30 A great piece of information about the CNF1 mechanism of function and modular structuration derived from a comparison with other well-known toxins such as the diphtheria toxin and the dermo-necrotic toxin.31 This homology-based hypothesis has been further corroborated by a series of in silico prediction stud-ies,1,29 mutagenesis experiments29,30 and interactome analy-sis25,28 so as to have the general idea of CNF1 structure as described above. Nevertheless, only the C-terminal catalytic domain of the protein (residues 720–1014) has been deeply studied from a structural point of view thanks to the work of Buetow et al. who have determined the X-ray crystal structure of this domain32 and its critical structural constrains for the catalyzed reaction.33
We think that such a detailed level of information should be extended to the remaining predominant part of the protein to achieve a clear understanding of its mechanism of action and its potential therapeutic use. In fact, there are still too many unsolved questions about CNF1 function. For example, why the binding with two different receptors is needed for the toxin internalization into eukaryotic cells? What are the roles of the two receptors? What does CNF1 conformational change experienced during endosomal acidification consist of? Which protease cleaves and provokes the release of the CNF1 catalytic domain in the cell cytosol? More impor-tantly, the use of CNF1 as a therapeutic for the treatment of neurological disorders demands an unbiased evaluation of its
intrinsic risks that can be fully contemplated only in the presence of data about the protein structure and function.
We think that the lack of structural studies about the full-length CNF1 issues from the difficulty to recombinantly pro-duce and purify it with high-yield in a simple way. Various research groups reported to still use the protocol established by Falzano et al.12which is long, complex, and leads to the harvest of insufficient amounts of protein. Particularly, this procedure consists of over ten steps including three ammo-nium sulphate precipitations alternated with dialysis and cen-trifugations, two ion exchange chromatographies, two gel filtrations and one hydrophobic exchange chromatography.12 Other groups testified a remarkable instability of the protein that was progressively proteolysed during its synthesis and difficult to separate from its degradation products;27,34 Schmidt’s research group routinely produces CNF1 as a GST-fused chimera with a reported yield of purified toxin of 1 mg per 4 liters of culture, but this yield remains unsatisfy-ing for a complete chemical-physical characterization.35
In this study, we propose a procedure aiming to increase both CNF1 production and purification yields, and to minimize its instability and degradation, for collecting sufficiently pure, concentrated and homogeneous CNF1.
Materials and Methods
Construction of pET40b-CNF1-H8 expression plasmid The cnf1 gene was inserted into the pET40b expression vector in three fragments. The central fragment of the gene (2150 bp) was obtained with BglII/NcoI double digestion of the pISS392 plasmid.1 The 50 part (190 bp) and the 30 part (740 bp) of the sequence were synthesised by Thermo Fisher Scientific and digested with NdeI/BglII and NcoI/XhoI restriction enzymes, respectively. The three fragments were then consecutively cloned into the pet40b plasmid between NdeI and XhoI restriction sites. The so obtained pET40b-cnf1-H8 vector allowed the production of CNF1-H8, a pro-tein encompassing the wild type 1014 residues fused to an 8xHis C-terminal tag encoded by the plasmid sequence. Expression and purification of CNF1-H8
recombinant protein
pET40b-cnf1-H8 vector was introduced into E. coli BL21(DE3) cells and different conditions were tested to assess CNF1-H8 expression. M9, LB, and TB36 and their variants were used as different growth media. Particularly, different carbon sources (glucose, glycerol, acetate -each added at 0.4– 1.0% w/v range-, and their combinations) were tested to define their effect on protein production quality.37–39IPTG induction was performed in different growth phases in a 0.1–1 mM range. Finally, 158C, 208C, and 378C were the explored temperatures during the growth and production phases. Once these different conditions had been tested, the production of the recombinant protein was routinely performed growing the
bacteria in LB medium in the presence of 50 mg/mL of kana-mycin at 208C. At OD6005 0.8, IPTG was supplemented to the final concentration of 0.5 mM and the temperature was shifted to 158C overnight.
After recombinant protein production in the chosen condi-tions, the bacterial cells were harvested by centrifugation (4500g, 48C, 30 min) and washed with phosphate-buffered saline (PBS). Then the pellets were resuspended in 5 mL/g wet cell weight of Lysis buffer (50 mM TrisHCl pH 8.0, 150 mM NaCl, 20 mM Imidazole, 15% v/v glycerol) supple-mented with the complete EDTA-free protease inhibitor cocktail (Roche). Afterwards, the suspensions were subjected to disruption by sonication (30 s cycles, 25% Amplitude with 30 s pauses between each cycle for a total process of 20 min at 48C) and the soluble and insoluble fractions were separated by centrifugation (13,000g, 48C, 45 min) and filtration.
The clarified lysates were applied to IMAC using a 1 mL HisTrap HP column (GE Healthcare) and an Akta purifier system (GE Healthcare). The bound fractions were washed and eluted using IMAC Wash buffer (50 mM TrisHCl pH 8.0, 150 mM NaCl, 50 mM Imidazole, 15% v/v glycerol) and IMAC Elution buffer (50 mM TrisHCl pH 8.0, 150 mM NaCl, 500 mM Imidazole, 15% v/v glycerol), respectively. The purity of the eluted fractions was further increased by a size exclusion chromatography (Superdex 200 pg, GE Healthcare) performed using 50 mM TrisHCl pH 8.0, 200 mM NaCl, 15% v/v glycerol as running buffer at 0.5 mL/min flow rate. At the end of the procedure, the recombinant proteins were dialyzed and concentrated (up to 10 mg/mL) in Storage buffer (50 mM TrisHCl pH 8.0, 200 mM NaCl, 30% v/v glycerol, 2 mM DTT) and preserved at 2208C.
Throughout the expression and purification procedures, CNF1-H8 preparations were separated by 10% SDS-PAGE and analyzed both by Coomassie staining and Western blot using a monoclonal anti-CNF1 antibody (NG8, Santa Cruz Biotechnology, 1:1,000 dilution) and a secondary anti-mouse antibody (1:10,000 dilution). Their concentration was assessed using the Bradford Protein Assay following the manufacturer’s instruction (Bio-Rad).
Dynamic light scattering (DLS)
The analytes frozen in Storage buffer at 10 mg/mL were diluted to 1 mg/mL in the chosen buffer and further equili-brated performing a diafiltration with Vivaspin 500 centrifugal concentrators (10 kDa, Sartorius) with two washes. In case of apparent precipitation, the suspensions were centrifuged and the soluble fractions were analyzed. The dispersity (Polydis-persion index, Pdi) and the particles size were determined at 258C performing four runs (15 measurements each) with a Zetasizer Nano ZS (Malvern Instruments Limited, Malvern, Worcestershire, UK). Table 1 presents the list of different
buffers and additives which were exploited to explore the protein stability in different storage conditions.
RhoA recombinant production, purification and SDS-urea PAGE mobility assay
RhoA was produced and purified following the protocol described by Self et al.40using the pGEX2T-N25RhoA vec-tor (kindly provided by E Lemichez, INSERM, Nice,
France). Briefly, E. coli TOP10 bearing the
pGEX2T-N25RhoA vector were induced with 0.1 mM IPTG at 208C for 16 h. After centrifugation, cells were lysed in 50 mM
TrisHCl pH 8.0, 50 mM NaCl, 5 mM MgCl2, 1 mM DTT.
GST-RhoA was purified using a GSTrap HP column (GE Healthcare) using 10 mM reduced glutathione during the elu-tion. After dialysis in Cleavage/Assay buffer (50 mM TrisHCl pH 8, 150 mM NaCl, 5 mM MgCl2, 2.5 mM CaCl2, 1 mM DTT), Thrombin (Novagen) was added to the sample (2U per mg of protein) for 20 h at 208C. GST was removed using a second glutathione affinity chromatography and RhoA was further purified by size exclusion chromatography (Superdex 75 pg, GE Healthcare). Around 50 mg of pure protein was collected from a 1 L culture and it was finally stored in the Cleavage/Assay buffer at 2208C at a concentra-tion of 1 mg/mL.
For in vitro mobility assays, recombinant RhoA protein (10 mM) was mixed with CNF1-H8 (0.2 mM) and incubated
in 50 mM TrisHCl pH 8.0, 150 mM NaCl, 5 mM MgCl2,
2.5 mM CaCl2, 1mM DTT for increasing times at 258C. The reaction was stopped by the addition of Laemmli buffer, the mixture was then subjected to SDS-PAGE on a 12.5% gel with 1M Urea and the proteins were detected by Coomassie staining. The deamidated form of RhoA (RhoA E63) gener-ated by CNF1-H8 action showed a delayed electrophoretic mobility over its unmodified form (RhoA Q63), as reported in refs.6,33
Multinucleation assay
Human epithelial HEp-2 cells (ATCCVR CCL23TM) were
grown in Dulbecco’s Modified Eagle’s Medium supple-mented with 10% fetal bovine serum (Flow Laboratories, Rockville, MD, USA), 5 mM L-glutamine, 100 U/mL peni-cillin and 100 lg/mL streptomycin. Cells were seeded at a density of 2 3 104cells/cm2in p24 well plates. Twenty-four hours after seeding, cells were exposed to serial dilution of CNF1-H8, starting from 1027M to 10216M. For the activity assay, the percentage of multinucleated cells was counted at each dilution and the minimal dose causing multinucleation in 50% of cells was identified.
Circular dichroism (CD)
CD spectra of CNF1-H8 have been collected by using a Jasco J-710 spectropolarimeter equipped with a Peltier thermostated cell holder (Model PTC-348WI) (Jasco, Easton, MD). The mean residue molar ellipticity, [h] in deg cm2 dmol21, has been calculated from:
h
½ 5 h½ obs3MRW= 10clð Þ
Where [h]obsis the measured ellipticity in degrees, MRW the mean residue molecular weight,c the protein concentration in g mL21and l is the optical path length in cm. Far-UV measure-ments (190–250 nm) have been performed by using a 0.1 cm
Table 1. List of Buffers and Additives Used During CNF1-H8 Stor-age and DLS Analysis at 48C
Tested buffers Tested additives 50 mM TrisHCl pH 8.0 5–30 % v/v Glycerol PBS pH 7.4 5–30 % v/v Glucose 100 mM carbonate buffer pH 9.3 20–500 mM NaCI 50 mM Bis-tris pH 6.5
8 mM Sodium Phosphate pH 6.6–7.8
cell and protein concentration 1 nM. Thermal unfolding curves have been recorded in the temperature mode at 209 nm between 20 and 708C. Temperature ramp 0.58C min21and 60 s equilibration time prior each record have been used. Data points have been collected thrice hence averaged. Data has been fitted by mean of Boltzmann sigmoid equation:
y5 A12A2 11eðx2x0Þ=dx1A2
Spectra deconvolution and secondary structure evaluation have been performed using Dichroweb server,41,42 by mean of either CONTIN43 or SELCON344 method (best matches achieved with reference sets 4 and 3 respectively).45,46
Results and Discussion
Production and purification of recombinant CNF1-H8 CNF1-H8 protein was produced in E. coli using a T7-based expression vector (pET40b). This system allowed the production of a His-tagged recombinant protein to simplify the purification procedures. Nevertheless, as already experi-enced by other research groups,34 CNF1-H8 overexpression in E. coli was characterized by the progressive release of some degradation products, probably due to proteolytic attack and its intrinsic instability. To face this issue, varia-bles like growth medium composition, temperature, inducer concentration and duration of induction were taken into account (see Materials and Methods section). The variation of either induction profile or medium composition did not significantly affect CNF1-H8 production quality (data not shown). Conversely, the selected temperature during the growth and induction phases seemed to be crucial. Figure 1 shows the SDS-PAGE (A) and Western blot (B) analysis of bacterial extracts collected after E. coli growth in the presence (lanes 1–3) and absence (lane 4) of IPTG induction.
In particular, lane 1 is representative of a bacterial culture grown at 208C until mid-exponential phase (OD6005 0.8) and induced overnight with 0.5 mM IPTG keeping the tem-perature constant. The condition represented by lane 2 just differed from the previous one in the use of a temperature downshift to 158C after induction. In lane 3, instead, a lysate obtained after a 6 h expression at 378C was loaded. The 116 kDa band characteristic of full-length CNF1-H8 was distin-guishable in induced lysates in both panels mainly when the expression was performed at low temperatures (highlighted with an arrow), while at 378C the soluble protein was mostly proteolyzed. The Western blot detection revealed the co-presence of some degradation products also when the expres-sion was performed at suboptimal temperatures (Figure 1B). Nonetheless, at 158C the intact protein resulted in the most prominent signal in comparison to its degradation products in the same lane (lane 2 in Figure 1), indicating that the tem-perature downshift was useful to stem CNF1-H8 proteolysis, which was significantly more evident in the other experimen-tal conditions. The expression performed at 208C gave the best result in terms of absolute recombinant protein produc-tion, as shown in lane 1 in Figure 1, but the degradation pro-cesses were also more significant. Considering that the NG8 Mab binds CNF1 C-terminal portion, the proteolyzed frag-ments highlighted in this Western blot are likely to preserve the 8xHis C-terminal tag and to be co-purified with the full-length protein during IMAC. In this way, it would have been quite complex to isolate the intact protein over its degrada-tion products, representing more than the 50% of the pro-duced CNF1-H8. That is why, we generally preferred to induce the production of the recombinant protein at 158C rather than 208C.
As reported in the Introduction section, the currently available protocols for CNF1 purification are either quite complex and time-consuming12or they lead to the collection of a final product whose quality and/or quantity is unsuitable
Figure 1. CNF1-H8 expression levels in E. coli BL21(DE3). Whole cell lysates were evaluated by SDS-PAGE (A) and Western blot analysis (B). 1, Induced cell lysate at 208C after overnight expression; 2, induced cell lysate after temperature downshift to 158C and overnight expression; 3, induced cell lysate at 378C after 6 h expression; 4, uninduced cell lysate. Full-length CNF1-H8 is indicated with an arrow.
for a fine biophysical characterization.27,34,35With the aim to overcome this limitation, we used an affinity chromatogra-phy to capture most of the recombinant protein followed by a size exclusion chromatography to increase the overall qual-ity of the collected fractions (Figure 2). In detail, the clari-fied lysate harvested after cells disruption was applied to immobilized metal affinity chromatography (IMAC) using a 1 mL HisTrap HP column. The bound proteins were washed and eluted with IMAC Wash buffer and IMAC Elution buffer, respectively. The eluted fractions (peak I in Figure 2A) were pooled and analyzed by both SDS-PAGE (lane 1 in Figure 2B) and Western blot (lane 2 in Figure 2B). The densitometric analysis of the Coomassie stained denaturant gel revealed a 75% purity of the target protein, while the Western blot showed that at least one main contaminant in the sample was a degraded form of CNF1-H8, as it was detected by NG8 Mab (the same major proteolyzed fragment shown in Figure 1B). To eliminate the remaining contami-nants, a size exclusion chromatography (Superdex 200 pg)
was performed on partially purified CNF1-H8. The chro-matogram presented two characteristic peaks, peak II eluted near the void limit of the column and peak III in the elution volume (Ve) range of 55–65 mL (Figure 2C). The SDS-PAGE related to peak II showed that a minor portion of the toxin was lost within the void volume of the column, possi-bly due to its partial aggregation (lane 1 in Figure 2D). Nev-ertheless, most of the CNF1-H8 eluted in peak III as shown by the SDS-PAGE and the Western blot analysis (lanes 2 and 3 in Figure 2D, respectively), which is compatible with the existence of a 116 kDa monomeric protein. Furthermore, the SDS-PAGE and Western blot performed on these col-lected fractions revealed an overall purity above 95% of the recombinant toxin in the final sample. In each lane of the Coomassie-stained gels and Western blot in Figures 2B,D, the amount of loaded proteins was normalized to 500 ng. Hence, the reader should refer to the chromatograms (Figures 2A,C) to have a concrete idea of the protein relative
abundance among the different samples. The whole
Figure 2. Two-steps purification of CNF1-H8. (A) Typical chromatogram of IMAC separation (HisTrap HP) of E. coli soluble cellular extracts after CNF1-H8 production. After flow-through discard, a washing step with mild imidazole concentration was followed by elution with 500 mM imidazole which led to the release of the target protein (Peak I). (B) Analysis of the eluted and pooled fractions after IMAC (Peak I). 1, SDS-PAGE; 2, Western blot. (C) The chromatogram of the second step purification of CNF1-H8 by size exclusion chromatography (Superdex 200 pg) displays two characteristic peaks, peak II representative of proteins released in the void volume of the column and peak III included in the Ve 55–65 mL range. (D) Analysis of the eluted and pooled fractions corresponding to Peaks II and III of the size exclusion chromatography. 1, PAGE related to Peak II; 2, SDS-PAGE of pure CNF1-H8 collected at the end of the purification procedure (Peak III); 3, pure CNF1 analyzed by Western blot (Peak III). 500 ng of total protein content was loaded in each lane of panels B and D.
purification protocol led to the collection of 7–8 mg of recombinant protein per liter of culture in highly pure form. In our opinion, this is a significant result as this yield undoubtedly outperforms the purification of 1 mg of protein from a 4 L culture reported in a previous publication.35 When the purification of CNF1-H8 was attempted on protein extracts collected after 208C expression, we could not achieve the same level of purity depicted in Figure 2D after the two chromatographic steps, because the about 80 kDa degradation product (Figures 1B and 2B) was entirely co-eluted after IMAC and the Superdex 200 pg was not able to completely separate the two proteins when their relative abundances were similar (data not shown). A third chromato-graphic step was needed to reach the same purity level, but this led to a considerable loss in terms of yield (data not shown).
Evaluation of stability and dispersity of purified CNF1-H8 A detailed and extensive biophysical characterization of CNF1 recombinant protein requires that the final protein for-mulation is characterized not only by a high purity level, but also by acceptable homogeneity, solubility, and stability over time. To assess these features, DLS experiments were rou-tinely performed to define the most suitable buffers to be used during the purification and storage of the protein.47 Fur-thermore, this procedure was also useful to define the behav-ior of CNF1-H8 when preserved in concentrated form (up to 10 mg/mL) for long periods of time.48This feature is of piv-otal importance as various structural and analytical techni-ques may require the use of highly concentrated and stable preparations. As consequence, the fulfillment of this goal could be crucial for a deeper structural characterization of CNF1.
First of all, DLS analysis was useful to define how the general quality of the protein preparations was increased in the two chromatographic steps. As shown in Figure 3A, CNF1-H8 collected after the IMAC was in a quite polydis-perse state (Pdi 0.501). The following size exclusion
chromatography led to the removal of complexed proteins characterized by a high hydrodynamic radius (Figure 3B) and to lower the Pdi of the protein solution to 0.275 (Figure 3C). This result confirms that peak II characteristic of the second step purification contained aggregated CNF1-H8 (Figure 2C). The almost monodisperse state of the final sam-ple was retained up to 10 days when the protein was pre-served in the storage buffer at 48C at a concentration of 10 mg/mL. At increased times of incubation, a slight increase of dispersity was observed (Figure 3D). These data confirm that our simple purification protocol permitted to collect CNF1-H8 in a pure, concentrated and stable form.
Once the target protein had been purified, DLS analysis was used as described inMaterials and Methods to study the protein behavior in different conditions mostly changing the buffers pH, ionic strength and testing additives known to sta-bilize proteins. Although CNF1-H8 seemed to be stable in a wide pH range (data not shown), the presence of salt and cosolvents (glycerol or glucose without a significant differ-ence) exhibited a pivotal role in the preservation of the pro-tein solubility and homogeneity. In fact, when CNF1-H8 was preserved in concentrated form (5–10 mg/mL) in 50 mM TrisHCl without additives and/or low NaCl concentration (up to 50 mM), it formed a turbid solution. This aggregation process was mainly affected by salt concentration, rather than other additives and this phenomenon was further con-firmed by DLS experiments performed on the soluble frac-tion of these suspensions. As shown in Table 2, the co-presence of 200 mM NaCl and 15% v/v glycerol allowed to keep the Pdi as low as 0.3 at 48C after 30 days storage. On the contrary, the single presence of glycerol with low NaCl content gave rise to a polydisperse suspension already after 24 h of incubation as indicated by a Pdi equal to 0.5. Finally, glycerol exhibited a minor still significant role in the preservation of CNF1-H8 stability, because the protein dis-solved in high salt buffer without other additives experienced a faster Pdi increase over time than the samples in 200 mM NaCl, 15% v/v glycerol.
Figure 3. Size distribution of CNF1-H8 solutions throughout its purification. (A) DLS analysis of CNF1-H8 purified by IMAC (Peak I). (B) DLS analysis of aggregated CNF1-H8 collected during size exclusion chromatography (Peak II). (C) DLS analysis of pure CNF1-H8 harvested at the end of the purification (Peak III). (D) Definition of CNF1-H8 stability when stored at 48C at a concentration of 10 mg/mL.
Thanks to these experiments, we could define that both a minimum NaCl concentration of 150 mM and the use of glycerol in the buffers could ameliorate the protein stability. For this reason, all the buffers used during the purification included these additives.
In vitro activity of CNF1-H8
Previous studies have demonstrated that the CNF1-catalyzed deamidation at position 63 in RhoA sequence can be monitored observing a typical electrophoretic shift in SDS-urea-PAGE.6 This qualitative assay was exploited to confirm that the purified full-length protein preserved its cat-alytic activity. Following a previously reported protocol, RhoA was recombinantly produced and purified.40 Then, RhoA was incubated with CNF1-H8 at 258C in a 50:1 molar ratio for increasing times. The deamidated form of RhoA (RhoA E63) generated by CNF1-H8 action was expected to become progressively predominant over its unmodified form (RhoA Q63) and to show a delayed electrophoretic mobility. In accordance with the previous results,33RhoA deamidation
was almost complete within 1 h as shown by the upward shift of RhoA characteristic band in an SDS-urea-PAGE experiment (Figure 4A). However, some kinetic differences were appreciated when CNF1-H8 was stored in different buf-fers. When the recombinant protein was preserved in high salt buffer with glycerol as a cosolvent, it provoked the con-version of about the 50% of the substrate in 10 min. Con-versely, the absence of either NaCl or glycerol in the storage buffer caused a partial inactivation of the protein, as visible in the first and second rows of Figure 4A. These results together with the DLS analysis reported in Table 2 demon-strate that the removal of salt and glycerol from the preser-vation buffer provoked CNF1-H8 aggregation and that the partially soluble aggregates that are formed partially preserve their catalytic activity.
To further investigate the CNF1-H8 activity, experiments on HEp-2 cells, an epithelial cell line representing the refer-ence cell line for CNF1 studies,49 were performed. HEp-2 cells were treated with CNF1-H8 at different doses and when analyzed by means of a phase contrast microscope, the typical effect of the wild type CNF1 was observed, i.e., cell spreading, multinucleation and ruffling at cell border (Figure 4B). The observed effects were dose-dependent as visible in Figure 4C. To obtain an index of the CNF1-H8 activity, HEp-2 cells were treated with doses of CNF1-H8 starting from 1027M and decreasing until 10216M. The lower dose at which CNF1-H8 induced multinucleation in 50% of the cells was identified at 10211M. The induction of the typical phenotypic effects on cultured cells indicates that the puri-fied toxin preserves its capability of being endocytosed, processed and released in the cell cytosol. Taken together, RhoA mobility assay and in vitro multinucleation assay con-firmed that the protein purified with our protocol exerts its biological functions.
Table 2. Monitoring of CNF1-H8 Pdi in Different Storage Conditions Tested additives Pdi after 24 h Pdi after 10 days Pdi after 30 days 15% v/v glycerol, 200 mM NaCl 0.275 0.277 0.323 15% v/v glycerol 0.505 0.591 0.617 200 mM NaCl 0.316 0.339 0.430
DLS analysis was performed on the soluble fraction of CNF1-H8 dissolved in 50 mM TrisHCl in the presence of 200 mM NaCl, 15% v/v glycerol, or both at regular times.
Figure 4. CNF1-H8 activity. (A) 200 nM CNF1-H8 and 10 mM RhoA were incubated at 258C for the indicated times and an SDS-urea-PAGE was run to distinguish the substrate (RhoA Q63, lower band) from the product (RhoA E63, upper band) gener-ated by the deamidation reaction. Before dilution in the Assay Buffer, CNF1-H8 was preserved in 50 mM TrisHCl buffer with 200 mM NaCl, or 15% v/v glycerol or both. (B) Phase contrast micrographs showing control Hep-2 cells and cells treated with different CNF1-H8 doses for 24 h. Bar represents 10 lm. (C) Percentage of multinucleated cells at different CNF1-H8 doses. Control: untreated cells.
